freedomintelligence / evaluation-of-chatgpt-on-information-extraction Goto Github PK

An Evaluation of ChatGPT on Information Extraction task, including Named Entity Recognition (NER), Relation Extraction (RE), Event Extraction (EE) and Aspect-based Sentiment Analysis (ABSA).

• Paper: Is Information Extraction Solved by ChatGPT? An Analysis of Performance, Evaluation Criteria, Robustness and Errors

ChatGPT has stimulated the research boom in the field of large language models. In this paper, we assess the capabilities of ChatGPT from four perspectives including Performance, Evaluation Criteria, Robustness and Error Types. Specifically, we first evaluate ChatGPT's performance on 17 datasets with 14 IE sub-tasks under the zero-shot, few-shot and chain-of-thought scenarios, and find a huge performance gap between ChatGPT and SOTA results. Next, we rethink this gap and propose a soft-matching strategy for evaluation to more accurately reflect ChatGPT's performance. Then, we analyze the robustness of ChatGPT on 14 IE sub-tasks, and find that: 1) ChatGPT rarely outputs invalid responses; 2) Irrelevant context and long-tail target types greatly affect ChatGPT's performance; 3) ChatGPT cannot understand well the subject-object relationships in RE task. Finally, we analyze the errors of ChatGPT, and find that "unannotated spans" is the most dominant error type. This raises concerns about the quality of annotated data, and indicates the possibility of annotating data with ChatGPT. The data and code are released at Github site.

All datasets, processed data and output result files are available at the google drive, except ACE04, ACE05 and TACRED raw datasets (for copyright reasons).

Download all the files, unzip them, and place them in the corresponding directories.

bash ./scripts/absa/eval.sh
bash ./scripts/ner/eval.sh
bash ./scripts/re/eval_rc.sh
bash ./scripts/re/eval_triplet.sh
bash ./scripts/ee/eval_trigger.sh
bash ./scripts/ee/eval_argument.sh
bash ./scripts/ee/eval_joint.sh

Before testing, you need to modify all --api_key and --result_file arguments in all *.sh scripts.

bash ./scripts/absa/report.sh
bash ./scripts/ner/report.sh
bash ./scripts/re/report_rc.sh
bash ./scripts/re/report_triplet.sh
bash ./scripts/ee/report_trigger.sh
bash ./scripts/ee/report_argument.sh
bash ./scripts/ee/report_joint.sh

By default, the metrics are calculated based on our output result files at Google Drive.

We will add the results and analysis of GPT-4.

@article{han2023-chatgpt-IE-evaluation,
  author       = {Ridong Han and
                  Tao Peng and
                  Chaohao Yang and
                  Benyou Wang and
                  Lu Liu and
                  Xiang Wan},
  title        = {Is Information Extraction Solved by ChatGPT? An Analysis of Performance, Evaluation Criteria, Robustness and Errors},
  journal      = {CoRR},
  volume       = {abs/2305.14450},
  year         = {2023},
  eprinttype   = {arXiv},
  eprint       = {2305.14450},
  url          = {https://doi.org/10.48550/arXiv.2305.14450},
  doi          = {10.48550/ARXIV.2305.14450},
}